AMR-ESVT: Efficient and Accuracy-Preserving Construction of Lightweight Effective Sound Velocity Tables via Adaptive Mesh Refinement

Abstract

Accurate compensation for acoustic ranging errors is essential for high-precision underwater positioning. However, existing effective sound velocity table (ESVT) methods remain constrained by the computational overhead of dense spatial grids and interpolation smoothing across localized acoustic transitions. Accordingly, a residual-driven adaptive mesh refinement (AMR) method, termed AMR-ESVT, is proposed for sparse ESVT construction and subsequent effective sound velocity (ESV) field reconstruction. The method integrates hash acceleration for first-arrival eigenray identification, local refinement driven by the second-order midpoint residual, and piecewise cubic Hermite interpolation within subdomains separated by indicated ESV discontinuity interfaces. Deep-ocean numerical simulations demonstrate minute-scale table construction and a reduction exceeding 74% in the number of stored ESVT entries relative to dense uniform tabulation, while maintaining low full-domain ESV reconstruction errors and approximately 98% reconstruction fidelity near indicated interfaces under a 0.1 m/s error threshold. Shallow-sea and deep-ocean trials further validate ESV reconstruction fidelity and propagation-time residual agreement with ray tracing under field measurement uncertainty. AMR-ESVT consequently reconciles rapid ESVT construction, lightweight tabular representation, and high-fidelity ESV reconstruction for sound speed compensation in high-precision underwater positioning.